Digital BW, The Print

> Dan,
> 
> it would make sense that a gs-to-RGB would stretch and twist that poor gs
> trying to fit it into whatever profile you have set for RGB (including a
> possible 
> gamma change). Doing channel copies in multichannel mode and then RGB
> should be the more profile-neutral way to go. Great pointer, thanks!
> 
> Antonis

I just now figured out what peculiar thing I was doing that made this
problem sporadic.  Normally I deal with grayscale images in a gamma 2.2
space.  If I convert to RGB I convert to Adobe RGB - a gamma 2.2 RGB space.
No problem.  The problem came when I got a file with a black to white
gradient in ColorMatch RGB.  Because I needed the overall image in grayscale
to test a B&W process I converted this to a 1.8 gamma gray space (so far so
good - histogram just fine).  Then, as part of an RGB process I was testing,
I converted it to my standard Adobe RGB space (2.2 gamma).  Poof - garbage
histogram and posterization in the shadows.  So I solved that by using a
Multichannel duplication process to get back to my Adobe RGB preference
rather than a gray to RGB conversion.

After a midnight "Aha!!!" experience, this morning I created a new file with
a black and white gradient in a 1.8 gamma gray space then converted it to a
2.2 gamma gray space.  The histogram exhibited the same gaps in the shadow
region as the conversion process I was using before.  Of course I had *just*
provided advice to a list member advising that grayscale files, if they are
to be converted to RGB, should go into an RGB space with the same gamma --
then I violated my own advice (I can only claim sleep deprivation as an
excuse).  At any rate I proved my point - even if unintentionally - in
grayscale,  a 1.8 to a 2.2 gamma conversion (and vice versa) is hardly
lossless, and that applies when the conversion is into a different gamma RGB
space.  Don't convert to a different gamma working space unless there is no
alternative and if you *have* to convert into a different gamma RGB space
use a multichannel Channel duplication process rather than a simple mode
change.

-- Dan Culbertson
so many years, so little time...